[experimental] Add support for live-in semantics of values in deopt bundles

[reames]

Summary

This is a first step towards supporting deopt value lowering and reporting entirely with the register allocator. I hope to build on this in the near future to support live-on-return semantics, but I have a use case which allows me to test and investigate code quality with just the live-in semantics so I've chosen to start there. For those curious, my use cases is our implementation of the "__llvm_deoptimize" function we bind to @llvm.deoptimize. I'm choosing not to hard code that fact in the patch and instead make it configurable via function attributes.

The basic approach here is modelled on what is done for the "Live In" values on stackmaps and patchpoints. (A secondary goal here is to remove one of the last barriers to merging the pseudo instructions.) We start by adding the operands directly to the STATEPOINT SDNode. Once we've lowered to MI, we extend the remat logic used by the register allocator to fold virtual register uses into StackMap::Indirect entries as needed. This does rely on the fact that the register allocator rematerializes. If it didn't along some code path, we could end up with more vregs than physical registers and fail to allocate.

Today, we *only* fold in the register allocator. This can create some weird effects when combined with arguments passed on the stack because we don't fold them appropriately. I have an idea how to fix that, but it needs this patch in place to work on that effectively. (There's some weird interaction with the scheduler as well, more investigation needed.)

My near term plan is to land this patch off-by-default, experiment in my local tree to identify any correctness issues and then start fixing codegen problems one by one as I find them. Once I have the live-in lowering fully working (both correctness and code quality), I'm hoping to move on to the live-on-return semantics. Note: I don't have any *known* miscompiles with this patch enabled, but I'm pretty sure I'll find at least a couple. Thus, the "experimental" tag and the fact it's off by default.

p.s. I'm not particularly attached to any of the attribute names or any of the other how aspects of how we decide to enable. If anyone has a better idea here, please make suggestions.

[sanjoy]

I don't have any deep problems with the patch, only some minor stylistic issues.

PS: This patch is one in a thousand. :)

include/llvm/IR/Statepoint.h

37 ↗

(On Diff #69589)

Doxygen comments here.

lib/CodeGen/SelectionDAG/StatepointLowering.cpp

38 ↗

(On Diff #69589)

I've never seen a cl::desc phrased as a question, perhaps phrase it as "If true, lower deopt bundles as live-in"?

edit: After reading the code, I'm not sure if this is needed. IIUC clients can already opt out of the new behavior by not specifying "deopt-lowering", so I'm not sure we need another knob.

455 ↗

(On Diff #69589)

Is this new assert related to this change? (It's fine to land with this change or separately, I'm just curious)

Also the convention for assertion failure messages in the rest of LLVM seems to be "Pointer without base(?|!)", so we should not deviate from that unless we need to.

463 ↗

(On Diff #69589)

s/conservative/conservatively/

465 ↗

(On Diff #69589)

Please add a small blurb on the difference between live-out and live-through. As we discussed in person, this can be subtle.

473 ↗

(On Diff #69589)

Not sure if hasBitSet is making things clearer. Does UseLiveInDeopt && (SI.StatepointFlags & StatepointFlags::DeoptLiveIn) compile?

476 ↗

(On Diff #69589)

Can this be

assert(is_contained(SI.Bases, V) implies is_contained(SI.Ptrs, V));
return is_contained(SI.Ptrs, V);

?

514 ↗

(On Diff #69589)

I'd add a /*LiveInOnly=*/ before the false, here and below.

lib/CodeGen/TargetInstrInfo.cpp

454 ↗

(On Diff #69589)

s/arguements/arguments/
s/args/arguments/

455 ↗

(On Diff #69589)

s/opers/Opers/

or

StartIdx = StatepointOpers(&MI).getVarIdx().

(I've fixed the existing violations).

lib/Transforms/Scalar/RewriteStatepointsForGC.cpp

1277 ↗

(On Diff #69589)

Not sure if the static is adding much here. I'd just stick with const char *DeoptLowering = "deopt-lowering". Also LLVM coding style associates the * with the variable name.

1340 ↗

(On Diff #69589)

Usually assertion failure messages are formatted as "Unsupported value!".

[reames]

Updated patch to follow shortly.

lib/CodeGen/SelectionDAG/StatepointLowering.cpp

455 ↗

(On Diff #69589)

It's new. It should already hold, but I noticed it while working on this.

473 ↗

(On Diff #69589)

Good catch. This was from I originally was trying to support live-out in the same patch and needed two bits to encode the options. Simplified. (In retrospect, I'd also picked a terrible name for the original purpose.)

476 ↗

(On Diff #69589)

No it can't. You can have a single use of the derived pointer after the call. This would produce a statepoint with a single gc.relocate with the pair (base, derived), but not (base, base). This is correct code generation for this case.

[reames]

Address Sanjoy's comments

[sanjoy]

lgtm with a couple of minor typos pointed out inline

lib/CodeGen/SelectionDAG/StatepointLowering.cpp

461 ↗

(On Diff #69732)

s/transative/transitive/

462 ↗

(On Diff #69732)

s/particularly/particular/

465 ↗

(On Diff #69732)

s/code generate/code generator/

llvm/trunk/include/llvm/IR/Statepoint.h

 
 namespace llvm {
 /// The statepoint intrinsic accepts a set of flags as its third argument.
 /// Valid values come out of this set.
 enum class StatepointFlags {
   None = 0,
   GCTransition = 1, ///< Indicates that this statepoint is a transition from
                     ///< GC-aware code to code that is not GC-aware.
+  /// Mark the deopt arguments associated with the statepoint as only being
+  /// "live-in". By default, deopt arguments are "live-through".  "live-through"
+  /// requires that they the value be live on entry, on exit, and at any point
+  /// during the call.  "live-in" only requires the value be available at the
+  /// start of the call.  In particular, "live-in" values can be placed in
+  /// unused argument registers or other non-callee saved registers.
+  DeoptLiveIn = 2,
 
-  MaskAll = GCTransition ///< A bitmask that includes all valid flags.
+  MaskAll = 3 ///< A bitmask that includes all valid flags.
 };
 
 class GCRelocateInst;
 class GCResultInst;
 class ImmutableStatepoint;
 
 bool isStatepoint(ImmutableCallSite CS);
 bool isStatepoint(const Value *V);

llvm/trunk/lib/CodeGen/SelectionDAG/StatepointLowering.cpp

#endif
 
   assert(Loc.getNode());
   return std::make_pair(Loc, Chain);
 }
 
 /// Lower a single value incoming to a statepoint node.  This value can be
 /// either a deopt value or a gc value, the handling is the same.  We special
 /// case constants and allocas, then fall back to spilling if required.
-static void lowerIncomingStatepointValue(SDValue Incoming,
+static void lowerIncomingStatepointValue(SDValue Incoming, bool LiveInOnly,
                                          SmallVectorImpl<SDValue> &Ops,
                                          SelectionDAGBuilder &Builder) {
   SDValue Chain = Builder.getRoot();
 
   if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Incoming)) {
     // If the original value was a constant, make sure it gets recorded as
     // such in the stackmap.  This is required so that the consumer can
     // parse any internal format to the deopt state.  It also handles null
     // pointers and other constant pointers in GC states.  Note the constant
     // vectors do not appear to actually hit this path and that anything larger
     // than an i64 value (not type!) will fail asserts here.
     pushStackMapConstant(Ops, Builder, C->getSExtValue());
   } else if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Incoming)) {
     // This handles allocas as arguments to the statepoint (this is only
     // really meaningful for a deopt value.  For GC, we'd be trying to
     // relocate the address of the alloca itself?)
     Ops.push_back(Builder.DAG.getTargetFrameIndex(FI->getIndex(),
                                                   Incoming.getValueType()));
+  } else if (LiveInOnly) {
+    // If this value is live in (not live-on-return, or live-through), we can
+    // treat it the same way patchpoint treats it's "live in" values.  We'll 
+    // end up folding some of these into stack references, but they'll be 
+    // handled by the register allocator.  Note that we do not have the notion
+    // of a late use so these values might be placed in registers which are 
+    // clobbered by the call.  This is fine for live-in.
+    Ops.push_back(Incoming);
   } else {
     // Otherwise, locate a spill slot and explicitly spill it so it
     // can be found by the runtime later.  We currently do not support
     // tracking values through callee saved registers to their eventual
     // spill location.  This would be a useful optimization, but would
     // need to be optional since it requires a lot of complexity on the
     // runtime side which not all would support.
     auto Res = spillIncomingStatepointValue(Incoming, Chain, Builder);
if (auto *GFI = Builder.GFI) {
     }
     for (const Value *V : SI.Ptrs) {
       auto Opt = S.isGCManagedPointer(V->getType()->getScalarType());
       if (Opt.hasValue()) {
         assert(Opt.getValue() &&
                "non gc managed derived pointer found in statepoint");
       }
     }
+    assert(SI.Bases.size() == SI.Ptrs.size() && "Pointer without base!");
   } else {
     assert(SI.Bases.empty() && "No gc specified, so cannot relocate pointers!");
     assert(SI.Ptrs.empty() && "No gc specified, so cannot relocate pointers!");
   }
 #endif
 
+  // Figure out what lowering strategy we're going to use for each part
+  // Note: Is is conservatively correct to lower both "live-in" and "live-out"
+  // as "live-through". A "live-through" variable is one which is "live-in",
+  // "live-out", and live throughout the lifetime of the call (i.e. we can find
+  // it from any PC within the transitive callee of the statepoint).  In
+  // particular, if the callee spills callee preserved registers we may not
+  // be able to find a value placed in that register during the call.  This is
+  // fine for live-out, but not for live-through.  If we were willing to make
+  // assumptions about the code generator producing the callee, we could
+  // potentially allow live-through values in callee saved registers.
+  const bool LiveInDeopt =
+    SI.StatepointFlags & (uint64_t)StatepointFlags::DeoptLiveIn;
+
+  auto isGCValue =[&](const Value *V) {
+    return is_contained(SI.Ptrs, V) || is_contained(SI.Bases, V);
+  };
+  
   // Before we actually start lowering (and allocating spill slots for values),
   // reserve any stack slots which we judge to be profitable to reuse for a
   // particular value.  This is purely an optimization over the code below and
   // doesn't change semantics at all.  It is important for performance that we
   // reserve slots for both deopt and gc values before lowering either.
   for (const Value *V : SI.DeoptState) {
+    if (!LiveInDeopt || isGCValue(V))
       reservePreviousStackSlotForValue(V, Builder);
   }
   for (unsigned i = 0; i < SI.Bases.size(); ++i) {
     reservePreviousStackSlotForValue(SI.Bases[i], Builder);
     reservePreviousStackSlotForValue(SI.Ptrs[i], Builder);
   }
 
   // First, prefix the list with the number of unique values to be
   // lowered.  Note that this is the number of *Values* not the
   // number of SDValues required to lower them.
   const int NumVMSArgs = SI.DeoptState.size();
   pushStackMapConstant(Ops, Builder, NumVMSArgs);
 
   // The vm state arguments are lowered in an opaque manner.  We do not know
   // what type of values are contained within.
   for (const Value *V : SI.DeoptState) {
     SDValue Incoming = Builder.getValue(V);
-    lowerIncomingStatepointValue(Incoming, Ops, Builder);
+    const bool LiveInValue = LiveInDeopt && !isGCValue(V);
+    lowerIncomingStatepointValue(Incoming, LiveInValue, Ops, Builder);
   }
 
   // Finally, go ahead and lower all the gc arguments.  There's no prefixed
   // length for this one.  After lowering, we'll have the base and pointer
   // arrays interwoven with each (lowered) base pointer immediately followed by
   // it's (lowered) derived pointer.  i.e
   // (base[0], ptr[0], base[1], ptr[1], ...)
   for (unsigned i = 0; i < SI.Bases.size(); ++i) {
     const Value *Base = SI.Bases[i];
-    lowerIncomingStatepointValue(Builder.getValue(Base), Ops, Builder);
+    lowerIncomingStatepointValue(Builder.getValue(Base), /*LiveInOnly*/ false,
+                                 Ops, Builder);
 
     const Value *Ptr = SI.Ptrs[i];
-    lowerIncomingStatepointValue(Builder.getValue(Ptr), Ops, Builder);
+    lowerIncomingStatepointValue(Builder.getValue(Ptr), /*LiveInOnly*/ false,
+                                 Ops, Builder);
   }
 
   // If there are any explicit spill slots passed to the statepoint, record
   // them, but otherwise do not do anything special.  These are user provided
   // allocas and give control over placement to the consumer.  In this case,
   // it is the contents of the slot which may get updated, not the pointer to
   // the alloca
   for (Value *V : SI.GCArgs) {

llvm/trunk/lib/CodeGen/TargetInstrInfo.cpp

case TargetOpcode::STACKMAP: {
     break;
   }
   case TargetOpcode::PATCHPOINT: {
     // For PatchPoint, the call args are not foldable (even if reported in the
     // stackmap e.g. via anyregcc).
     StartIdx = PatchPointOpers(&MI).getVarIdx();
     break;
   }
+  case TargetOpcode::STATEPOINT: {
+    // For statepoints, fold deopt and gc arguments, but not call arguments.
+    StartIdx = StatepointOpers(&MI).getVarIdx();
+    break;
+  }
   default:
     llvm_unreachable("unexpected stackmap opcode");
   }
 
   // Return false if any operands requested for folding are not foldable (not
   // part of the stackmap's live values).
   for (unsigned Op : Ops) {
     if (Op < StartIdx)
MachineInstr *TargetInstrInfo::foldMemoryOperand(MachineInstr &MI,
 
   MachineBasicBlock *MBB = MI.getParent();
   assert(MBB && "foldMemoryOperand needs an inserted instruction");
   MachineFunction &MF = *MBB->getParent();
 
   MachineInstr *NewMI = nullptr;
 
   if (MI.getOpcode() == TargetOpcode::STACKMAP ||
-      MI.getOpcode() == TargetOpcode::PATCHPOINT) {
+      MI.getOpcode() == TargetOpcode::PATCHPOINT ||
+      MI.getOpcode() == TargetOpcode::STATEPOINT) {
     // Fold stackmap/patchpoint.
     NewMI = foldPatchpoint(MF, MI, Ops, FI, *this);
     if (NewMI)
       MBB->insert(MI, NewMI);
   } else {
     // Ask the target to do the actual folding.
     NewMI = foldMemoryOperandImpl(MF, MI, Ops, MI, FI, LIS);
   }
#endif
   MachineBasicBlock &MBB = *MI.getParent();
   MachineFunction &MF = *MBB.getParent();
 
   // Ask the target to do the actual folding.
   MachineInstr *NewMI = nullptr;
   int FrameIndex = 0;
 
   if ((MI.getOpcode() == TargetOpcode::STACKMAP ||
-       MI.getOpcode() == TargetOpcode::PATCHPOINT) &&
+       MI.getOpcode() == TargetOpcode::PATCHPOINT ||
+       MI.getOpcode() == TargetOpcode::STATEPOINT) &&
       isLoadFromStackSlot(LoadMI, FrameIndex)) {
     // Fold stackmap/patchpoint.
     NewMI = foldPatchpoint(MF, MI, Ops, FrameIndex, *this);
     if (NewMI)
       NewMI = &*MBB.insert(MI, NewMI);
   } else {
     // Ask the target to do the actual folding.
     NewMI = foldMemoryOperandImpl(MF, MI, Ops, MI, LoadMI, LIS);

llvm/trunk/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp

if (IsDeoptimize) {
       RI->eraseFromParent();
     }
 
     OldI->eraseFromParent();
   }
 };
 }
 
+static StringRef getDeoptLowering(CallSite CS) {
+  const char *DeoptLowering = "deopt-lowering";
+  if (CS.hasFnAttr(DeoptLowering)) {
+    // FIXME: CallSite has a *really* confusing interface around attributes
+    // with values.  
+    const AttributeSet &CSAS = CS.getAttributes();
+    if (CSAS.hasAttribute(AttributeSet::FunctionIndex,
+                          DeoptLowering))
+      return CSAS.getAttribute(AttributeSet::FunctionIndex,
+                               DeoptLowering).getValueAsString();
+    Function *F = CS.getCalledFunction();
+    assert(F && F->hasFnAttribute(DeoptLowering));
+    return F->getFnAttribute(DeoptLowering).getValueAsString();
+  }
+  return "live-through";
+}
+    
+
 static void
 makeStatepointExplicitImpl(const CallSite CS, /* to replace */
                            const SmallVectorImpl<Value *> &BasePtrs,
                            const SmallVectorImpl<Value *> &LiveVariables,
                            PartiallyConstructedSafepointRecord &Result,
                            std::vector<DeferredReplacement> &Replacements) {
   assert(BasePtrs.size() == LiveVariables.size());
 
makeStatepointExplicitImpl(const CallSite CS, /* to replace */
 
   StatepointDirectives SD =
       parseStatepointDirectivesFromAttrs(CS.getAttributes());
   if (SD.NumPatchBytes)
     NumPatchBytes = *SD.NumPatchBytes;
   if (SD.StatepointID)
     StatepointID = *SD.StatepointID;
 
+  // Pass through the requested lowering if any.  The default is live-through.
+  StringRef DeoptLowering = getDeoptLowering(CS);
+  if (DeoptLowering.equals("live-in"))
+    Flags |= uint32_t(StatepointFlags::DeoptLiveIn);
+  else {
+    assert(DeoptLowering.equals("live-through") && "Unsupported value!");
+  }
+
   Value *CallTarget = CS.getCalledValue();
   if (Function *F = dyn_cast<Function>(CallTarget)) {
     if (F->getIntrinsicID() == Intrinsic::experimental_deoptimize) {
       // Calls to llvm.experimental.deoptimize are lowered to calls to the
       // __llvm_deoptimize symbol.  We want to resolve this now, since the
       // verifier does not allow taking the address of an intrinsic function.
 
       SmallVector<Type *, 8> DomainTy;

llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll

+; RUN: llc -O3 < %s | FileCheck %s
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.11.0"
+
+declare void @bar() #0
+declare void @baz()
+
+define void @test1(i32 %a) gc "statepoint-example" {
+entry:
+; We expect the argument to be passed in an extra register to bar
+; CHECK-LABEL: test1
+; CHECK:       pushq  %rax
+; CHECK-NEXT: Ltmp0:
+; CHECK-NEXT:  .cfi_def_cfa_offset 16
+; CHECK-NEXT: callq _bar
+  %statepoint_token1 = call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 1, i32 %a)
+  ret void
+}
+
+define void @test2(i32 %a, i32 %b) gc "statepoint-example" {
+entry:
+; Because the first call clobbers esi, we have to move the values into
+; new registers.  Note that they stay in the registers for both calls.
+; CHECK-LABEL: @test2
+; CHECK:       movl %esi, %ebx
+; CHECK-NEXT:  movl %edi, %ebp
+; CHECK-NEXT: callq _bar
+  call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 2, i32 %a, i32 %b)
+  call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 2, i32 %b, i32 %a)
+  ret void
+}
+
+define void @test3(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h, i32 %i) gc "statepoint-example" {
+entry:
+; TODO: We should have folded the reload into the statepoint.
+; CHECK-LABEL: @test3
+; CHECK:        movl  32(%rsp), %r10d
+; CHECK-NEXT:   movl  24(%rsp), %r11d
+; CHECK-NEXT:   movl  16(%rsp), %eax
+; CHECK-NEXT:   callq _bar
+  %statepoint_token1 = call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 9, i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h, i32 %i)
+  ret void
+}
+
+; This case just confirms that we don't crash when given more live values
+; than registers.  This is a case where we *have* to use a stack slot.
+define void @test4(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h, i32 %i, i32 %j, i32 %k, i32 %l, i32 %m, i32 %n, i32 %o, i32 %p, i32 %q, i32 %r, i32 %s, i32 %t, i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z) gc "statepoint-example" {
+entry:
+; TODO: We should have folded the reload into the statepoint.  
+; CHECK-LABEL: test4
+; CHECK: pushq %r15
+; CHECK: pushq %r14
+; CHECK: pushq %r13
+; CHECK: pushq %r12
+; CHECK: pushq %rbx
+; CHECK: pushq %rax
+; CHECK:        movl  128(%rsp), %r13d
+; CHECK-NEXT:   movl  120(%rsp), %r12d
+; CHECK-NEXT:   movl  112(%rsp), %r15d
+; CHECK-NEXT:   movl  104(%rsp), %r14d
+; CHECK-NEXT:   movl  96(%rsp), %ebp
+; CHECK-NEXT:   movl  88(%rsp), %ebx
+; CHECK-NEXT:   movl  80(%rsp), %r11d
+; CHECK-NEXT:   movl  72(%rsp), %r10d
+; CHECK-NEXT:   movl  64(%rsp), %eax
+; CHECK-NEXT:   callq _bar
+  %statepoint_token1 = call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 26, i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h, i32 %i, i32 %j, i32 %k, i32 %l, i32 %m, i32 %n, i32 %o, i32 %p, i32 %q, i32 %r, i32 %s, i32 %t, i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z)
+  ret void
+}
+
+; A live-through gc-value must be spilled even if it is also a live-in deopt
+; value.  For live-in, we could technically report the register copy, but from
+; a code quality perspective it's better to reuse the required stack slot so 
+; as to put less stress on the register allocator for no benefit.
+define  i32 addrspace(1)* @test5(i32 %a, i32 addrspace(1)* %p) gc "statepoint-example" {
+entry:
+; CHECK-LABEL: test5
+; CHECK:        movq  %rsi, (%rsp)
+; CHECK-NEXT:   callq _bar
+  %token = call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 1, i32 %a, i32 addrspace(1)* %p, i32 addrspace(1)* %p)
+  %p2 = call i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(token %token,  i32 9, i32 9)
+  ret i32 addrspace(1)* %p2
+}
+
+; Show the interaction of live-through spilling followed by live-in.
+define void @test6(i32 %a) gc "statepoint-example" {
+entry:
+; TODO: We could have reused the previous spill slot at zero additional cost.
+; CHECK-LABEL: test6
+; CHECK:        movl %edi, %ebx
+; CHECK:        movl %ebx, 12(%rsp)
+; CHECK-NEXT:   callq _baz
+; CHECK-NEXT:  Ltmp30:
+; CHECK-NEXT:   callq _bar
+  call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @baz, i32 0, i32 0, i32 0, i32 1, i32 %a)
+  call token (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 1, i32 %a)
+  ret void
+}
+
+
+; CHECK: Ltmp1-_test1
+; CHECK:      .byte 1
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .short  5
+; CHECK-NEXT: .long 0
+
+; CHECK: Ltmp7-_test2
+; CHECK:      .byte 1
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .short  6
+; CHECK-NEXT: .long 0
+; CHECK:      .byte 1
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .short  3
+; CHECK-NEXT: .long 0
+; CHECK: Ltmp8-_test2
+; CHECK:      .byte 1
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .short  3
+; CHECK-NEXT: .long 0
+; CHECK:      .byte 1
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .short  6
+; CHECK-NEXT: .long 0
+
+declare token @llvm.experimental.gc.statepoint.p0f_isVoidf(i64, i32, void ()*, i32, i32, ...)
+declare i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(token, i32, i32)
+
+
+attributes #0 = { "deopt-lowering"="live-in" }
+attributes #1 = { "deopt-lowering"="live-through" }

llvm/trunk/test/Transforms/RewriteStatepointsForGC/deopt-lowering-attrs.ll

+; RUN: opt -rewrite-statepoints-for-gc -S < %s | FileCheck %s
+; Check that the "deopt-lowering" function attribute gets transcoded into
+; flags on the resulting statepoint
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.11.0"
+
+declare void @foo()
+declare void @bar() "deopt-lowering"="live-in"
+declare void @baz() "deopt-lowering"="live-through"
+
+define void @test1() gc "statepoint-example" {
+; CHECK-LABEL: @test1(
+; CHECK: @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @foo, i32 0, i32 0, i32 0, i32 1, i32 57)
+; CHECK: @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @bar, i32 0, i32 2, i32 0, i32 1, i32 42)
+; CHECK: @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 2882400000, i32 0, void ()* @baz, i32 0, i32 0, i32 0, i32 1, i32 13)
+
+entry:
+  call void @foo() [ "deopt"(i32 57) ]
+  call void @bar() [ "deopt"(i32 42) ]
+  call void @baz() [ "deopt"(i32 13) ]
+  ret void
+}